What problem does it solve? Responding to peer review comments requires mapping each criticism to actual experimental evidence, avoiding fabrication, and ensuring no concern is left unanswered. This Skill automates that process by parsing reviews, atomizing concerns, checking evidence sufficiency against a research wiki, and drafting traceable rebuttals. ## Core Features & Use Cases - Concern Atomization: Splits reviewer weaknesses and questions into atomic concerns with Rvx-Cy IDs, classified by type (evidence, method, missing, clarity, scope, novelty) and severity. - Evidence Mapping: Links each concern to wiki claims and experiments, judging evidence as sufficient, partial, insufficient, or contradicted to select the right response strategy. - LLM Stress-Test: Simulates skeptical reviewer follow-ups via a Review LLM, scoring each response 1-5 and revising weak answers across up to two rounds. - Use Case: After receiving three reviewer reports for an ICLR submission, run the Skill on the review files to produce a formal plain-text rebuttal for the submission system plus a rich-text analysis with evidence gap tables and suggested follow-up experiments. ## Quick Start Run the rebuttal skill on raw/reviews/reviewer1.txt with paper slug my-paper to generate formal and rich-text rebuttals in wiki/outputs/.
